Update README.md
Browse files
README.md
CHANGED
|
@@ -1,3 +1,10 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
Pretrained model based on [microsoft/deberta-v3-base](https://huggingface.co/microsoft/deberta-v3-base) with further mathematical pre-training.
|
| 2 |
|
| 3 |
Compared to deberta-v3-base, 300 additional mathematical LaTeX tokens have been added before the mathematical pre-training. As this additional pre-training used NSP-like tasks, a pooling layer has been added to the model (`bias` and `weight`). If you don't need this pooling layer, just use the standard transformers DeBERTa model. If you want to use the additional pooling layer like the BERT one, a wrapper class like the following may be used:
|
|
|
|
| 1 |
+
---
|
| 2 |
+
datasets:
|
| 3 |
+
- ddrg/named_math_formulas
|
| 4 |
+
- ddrg/math_formula_retrieval
|
| 5 |
+
- ddrg/math_formulas
|
| 6 |
+
- ddrg/math_text
|
| 7 |
+
---
|
| 8 |
Pretrained model based on [microsoft/deberta-v3-base](https://huggingface.co/microsoft/deberta-v3-base) with further mathematical pre-training.
|
| 9 |
|
| 10 |
Compared to deberta-v3-base, 300 additional mathematical LaTeX tokens have been added before the mathematical pre-training. As this additional pre-training used NSP-like tasks, a pooling layer has been added to the model (`bias` and `weight`). If you don't need this pooling layer, just use the standard transformers DeBERTa model. If you want to use the additional pooling layer like the BERT one, a wrapper class like the following may be used:
|